Long-form, practical, and specific: Claude Code workflows, agent architecture, and prompting technique — the write-ups we wished existed while building AgentsKit.
Not a directory dump — a ranked shortlist of the MCP servers worth connecting, why Context7 is the first one to install, which ones duplicate what Claude Code already does, and why four to six beats forty.
9 min read · Read →Plugins bundle skills, agents, commands, hooks, and MCP servers into one versioned install. Which ones are worth it, the sixty-second test for judging any plugin, and the three kinds you should never install.
9 min read · Read →Claude Code, Cursor, Codex, Copilot, Cline, Aider and more — ranked on completed work rather than demos, with a pick-by-job table and the configuration variable that decides most of the difference.
10 min read · Read →Hooks are the only layer of your setup that can't be ignored. The nine worth having — auto-format, dangerous-command blocking, secret guards, test runs, notifications — plus which event to use for what.
9 min read · Read →Seven tools that cover the whole job instead of fifty that overlap: production, keyword data, research, design, video, lifecycle, and automation — with a clear #1 and the rule that keeps AI content from becoming slop.
9 min read · Read →Worked examples of Claude skills — commit messages, tests, PDFs, brand voice, debugging, and project conventions — with the anatomy of a SKILL.md and the two rules that decide whether a skill ever fires.
8 min read · Read →Code review, tests, debugging, refactoring, security, migrations, git hygiene, and your own conventions — what goes inside each coding skill, and when to use a skill instead of CLAUDE.md or a subagent.
9 min read · Read →There is no single app store for Claude skills — there are many marketplaces, GitHub repos, and curated kits. How plugin marketplaces work, how to vet a skill in thirty seconds, and what is actually worth paying for.
8 min read · Read →Reading a PDF is not the same as working with one. How a PDF skill bundles scripts to extract tables, fill forms, merge, split, OCR scans, and generate documents — and the pattern worth stealing for every other skill.
8 min read · Read →Almost every Claude skill lives on GitHub — unreviewed and sorted by nothing useful. Where the good ones are, how to copy a skill folder into place, how to judge a repo in a minute, and the safety rules people skip.
8 min read · Read →Fluency is not rigor. How a research skill encodes a search protocol, a sourcing rule with teeth, and permission to find nothing — plus the five research skills worth having, for academics and analysts alike.
8 min read · Read →How skills work (progressive disclosure), how they differ from CLAUDE.md, subagents, and MCP, the let-the-model-decide-let-the-code-execute pattern, and how to roll skills out across a team via git.
9 min read · Read →Reading a fluent explanation feels like learning and mostly isn't. Six study skills — Socratic tutoring, active recall, the Feynman technique, spaced repetition, worked examples, and exam simulation — plus how to write one in five minutes.
8 min read · Read →Why most product posts get removed or shadow-banned on Reddit — and a copy-paste Claude prompt that interviews you about your product, karma, and goal, then writes a post that reads like a real person, not an ad.
8 min read · Read →The four ways to extend Claude Code, explained side by side: what skills, subagents, slash commands, and MCP servers each do, how they differ, and a decision table for picking the right one.
9 min read · Read →A curated starter set of Claude Code subagents — code reviewer, debugger, security auditor, orchestrator, and more — with the reasoning behind each pick and what makes a subagent actually good.
9 min read · Read →Why most Claude Code skills never fire — and how to fix it. How skill triggering works, the description formula, the five most common failure causes, and how to test a skill like code.
9 min read · Read →What to put in CLAUDE.md (and what to leave out): a copy-paste template, the under-200-lines rule, what earns a line, and the habits that keep the file working as your project grows.
9 min read · Read →Fifteen ready-to-use custom slash commands for Claude Code — review, commit, debug, changelog, dependency audits, and more — plus the 60-second syntax primer and the patterns behind good commands.
10 min read · Read →A troubleshooting guide for subagents that never get used or return garbage: delegation-tuned descriptions, context bleed in both directions, tool scoping, and a 5-minute diagnostic.
9 min read · Read →Sequential, parallel, and fan-out orchestration for Claude Code agents — when each pattern earns its cost, seven rules for keeping token spend sane, and a worked three-agent example.
10 min read · Read →Claude Code isn't just for developers. How marketers use it for SEO content pipelines, campaign assets, and reporting — working on real files, no code required, with concrete prompts.
9 min read · Read →Why $300 SaaS boilerplates stopped making sense: locked stacks, dead code, and rot — and the agent-first build sequence that replaces them with code generated for your stack.
9 min read · Read →An opinionated shortlist of Claude Code skills worth installing in 2026 — code review, security, documents, SEO, and more — plus how to judge any skill in thirty seconds and install safely.
9 min read · Read →A practical, no-fluff guide to Claude Code: installation, your first session, CLAUDE.md, permissions, skills and slash commands — and the workflow habits that separate power users from tourists.
9 min read · Read →What an AI agent actually is, when you need one (and when you don't), the agent loop, tool use, MCP, and how to build your first working agent with the Claude API and Agent SDK.
12 min read · Read →The ten techniques that consistently improve Claude's output: context front-loading, examples over adjectives, XML structure, chain-of-thought, role prompting, output prefills, and more — each with a before/after.
10 min read · Read →Seven copy-paste Claude prompts for investing research: reading annual reports, stress-testing a thesis, decoding earnings calls, and building a decision journal. Educational, not financial advice.
8 min read · Read →The prompt library has 50 field-tested Claude prompts across coding, writing, data, business, and marketing — each with the reasoning behind it.